Thagard, P. (2008). How molecules matter to mental computation. In P. Thagard (Ed.) Hot Thought: Mechanisms and applications of emotional cognition pp 115-131. Cambridge, MA: MIT Press.
In this chapter, Thagard argues that computational models of cognition need to consider the influence of neuromodulators at the molecular level. He argues for understanding processes in the brain as the result of both chemical and electrical activity. He goes on to point out that many of the chemical influences on synaptic activation occur as the result of activity of cells far removed from the local neural network, such as when the release of hormones influence distant synaptic firing. Much of the chapter goes into technical details of how chemicals such as hormones influence the action of neurotransmitters and synaptic activity, but the overall thesis is that if cognitive scientist are to construct accurate computational models, they must take into consideration the effects of chemical processes on the activation of these models, rather than approach them as if they were electrical computers. The effects of chemical processes, he argues, are even more important to consider in computational models of cognition as the evidence for the role of emotion in cognition mounts. Evidence already exists for the effects of hormones or other neuromodulators on emotion; therefore, these same chemical reactions ultimately affect cognition. He points out that while it may not be necessary to only consider systems at the molecular level, knowledge about molecular processes should be considered one type of map that is useful for certain levels of analyses – he presents the example that while a large map of Europe is useful for locating Switzerland as north of Italy, another type of map, more fine grained, is useful for navigating the terrain of the Swiss Alps. I would tend to agree that if a truly holistic account of cognition is to be developed, consideration of the molecular processes contributing to patterns of neural activation would only serve to enrich this development. While it might not serve cognitive science for researchers to try to attempt the formidable task of becoming expert in all levels of analysis, integrating findings from molecular and network levels of analyses would likely strengthen the understanding of cognition at both of these levels and strengthen a more holistic understanding.